Any BSEL done yet for mobile Penryn?

[aicjofs]

My internet search skills must have went down hill as I can't seem to find much if anything on this.

I have a T6400 here that I was going to play with on a GM45 chipset. 10x multi, 800FSB. Seems like a BSEL to 1066 would turn it into a T9550 with less L2 cache.

I know from previous experience on Merom that the BSEL mods on Intel chipsets were locking the multi at 6x then which pretty much defeated any overclocking attempts, although other chipsets(i.e AMD/ATI) the mod would work correctly. I had a modded BIOS I made for Merom that I wanted to try, and then sold the laptop.

I thought I'd ask around before I go ripping things apart(laptops are so much more difficult then a desktop .. ), but perhaps I'll go into the unknown alone...if anyone has seen or heard about it link me

This is the only thing even remotely close and it's not that close... http://www.xtremesystems.org/forums/....php?p=3149607

Cheers

[aicjofs]

So I decided to try it. Seems the same with Penryn and GM45, as Merom was before. It's sucessful 200FSB->266FSB but stuck multi at 6x. Shame I don't have the tools to try a BIOS mod for this BIOS.

Someone mentioned they got around this in Linux...need to look into that next.

[The Coolest]

You may want to try RMClock. It can adjust multiplier and voltage depending on CPU load automatically

[aicjofs]

I have tried a multitude of things for this attempt and the previous Merom try in RMClock, CrystalCPUID, etc it seems locked at a chipset level, which is weird since I find little detail of anything it might be in the whitepaper for 4 series. http://www.intel.com/Assets/PDF/datasheet/320122.pdf


Again since history, and practice around the internet on Merom have it only taking place on an Intel chipset, where on other chipsets the lock is not in place. I didn't find much difinitive looking through patents either. For example http://www.freepatentsonline.com/y20...ng&stemming=on defiently addresses the BSEL "hardwire mod" but the solution is on the CPU in which case if this patent was what was used it wouldn't work on other chipsets, or even the desktop CPU's for that matter.

@MeltedDuron

It wasn't any faster really, since the gain in FSB was offset by 400Mhz loss in CPU speed due to locked 6x multi. I suppose battery life would have been better.


I never remember to take pictures when I do things, last mod I remembered to take pictures for was when we working on the 7900GT voltage mod...
That makes it harder to work up a good guide.

There are a few guides out there for Merom socket P which isn't much different then what we have here. But basically for your T6400 to go from 200FSB to 266FSB reference this document http://download.intel.com/design/mob...s/32012001.pdf

Beside the tools to take the laptop apart I use(hopefully just a fine phillips or torx and a flat head to do any prying to pop tabs)

Twizzers
Old IDE ribbon cable
Xacto knife

Basically just want a single piece of wire out of the IDE cable. Use the Xacto to cut one of the wires off the IDE cable, only need an inch, take the insulation off, and there will be a small bundle of wire. Now cut/extract a single wire from the bundle. It's length needs to be about as much as to make a "U" shape and go from one hole in the socket to another each leg of the U going in about pin depth into the socket.

So from the above document we have BSEL diagram



The T6400 had 200FSB and therefore L H L on BSEL pins, to get 266FSB we need to pull BSEL[1] to ground and get L L L. There is a VSS pin right next to BSEL[1] so this is where we will put the loop of wire.

Now referencing the CPU diagram from the document. This is a top down view so in effect we are looking at the socket in this picture. So looking at the below as if it were the socket itself place the loop of wire where the red mark is. It's important not to drop the wire as it is very small and Murphy Law seems to be at work if you drop it. Once the loop is in place reinstall the CPU. It won't drop in as easy since the wire will be fighting the pin for space in the hole. I usually try to install that edge of the CPU first. It's more of a commom sense thing, you'll need a little pressure to get it in but shouldn't require brute force. Now make sure the CPU is sitting flush in the socket, you don't want it raised on one edge causing poor heatsink contact. Then reassemble everything.



But as I said before this is an excessive amount of work if the multi stays locked at 6x, there is little benefit except the elevated FSB, but lower overall CPU speed. I have also found that conductive paint will do the job just fine as well depending on CPU, job, etc, however it works best when there are edge pins involved(i.e Athlon 64 VID pins are on the side), and you can "draw" along the edge of the cpu(also easier to remove if you want when it's on the edge). Since this mod is moving towards the center of cpu I find the wire trick the better option in this case, but feel free to try conductive paint if you wish.

[saaya]

yeah, found the same problem on my vaio centrino2, im on 266fsb already but want 333 or 400 if not more... i mean come on! this is P45 we are talking about here...
i wonder if disabling eist would help... maybe hardmod it somehow to force eist to not work?

[ExodusC]

I have a T6400 in my Acer Aspire 6930, and I've love to be able to push it a little further if possible. I hope something comes of this.

[aicjofs]

@saaya

haha, perhaps I have my sights set too low.

Well assuming the BIOS/chipset recognize the strap the folowing would logically be what we were looking at from other Intel whitepapers.



I know the 940 and 943 chipsets officially only supported 100 and 133 but would work quite well when a 166 was dropped in and work without issue, so perhaps we could get lucky as well with GM45. 400 might be asking for alot but would be perfect as far as ony one pin would need modded and you would actually get a bump in overall CPU speed even if you were stuck at 6x, of course some of the power saving features in the laptop would then be lost. Since dynamic FSB throttling is a 50% reference thing it still may work.

In the old days you isolated a pin to bring it high but it seems I read a few years back about using VCC to bring the pin high(would save the nail polish trick or the awful snapping the pin off), there is VCC close enough to do it, I need to reread for whatever reason I have been lucky and haven't had to go "high" for awhile.

It does seem that it is EIST/C1E at work, multiplier and voltage drop to lowest level in the manner they function. Yet is still different because it can't be disabled through the normal means. The only thing I puzzle over again is that EIST or C1E is enabled via BIOS and software controlled via CPU register(hence why it normally can be operated by RMClock for example) and the only way this lock has been overidden in these BSEL mods on mobile platform in the past is on non-Intel chipsets. Maybe it's just actually in the BIOS code that would be specifc to an Intel chipset. Usually Intel makes a base(default) BIOS code and manufactures adjust as necessary. A comparision of chip rated speed versus actual at initization time. Not like I'm breaking ground here, this was talked to death on a few forums for Merom.

The only other thing I have ever read about a stuck 6x is when the BIOS didn't recognize the chip, no overclocking in play just a non recognized chip. The T6400 doesn't have an identical 266 chip so that may be out. Not sure I was always stuck with the thought it was a BIOS issue related specifically to the Intel chipset written BIOS.

[saaya]

are you going to try this 400 strap mod then? let me know if you do!
it might actually work...

about the bios, well eist has been acting funny a lot when unsupported cpus were used, i still remember playing with a yorkfield on an i680 board that booted with stock multiplier but once it was in windows ran with the multiplier we set in bios...

ive seen the same with dothan and other mobile cpus used on desktop boards with the asus ct479 adapter, it seems that eist boots with the default multiplier and if it doesnt know the default multiplier then it will run the lowest supported multiplier. same for windows, it will run the lowest supported multiplier and then switch to the highest supported one if its under load.

so i assume that the registers that store the information of the default ie highest supported/allowed multiplier are not working right... its possible that intel uses a diferent set of registers for mobile cpus to store these values compared to the desktop cpus...

in the end i came to the conclusion that the best way to fix this was to increase the fsb in windows, since we really dont care about the cpu speed during post/bootup... sure it might reduce boot time by a second or so, but whos booting their laptops nowadays anywhow? its all about s3/s4 now isnt it?
the 266 strap is the tightest one, best perf per clock, at least on p35 and x38/x48 it was... for fsb ocing 266 was limited compared to 333, but it clocked almost the same, and def high enough for what we are looking for.

so i came to the conclusion what we really need is PLL/clockgen mods, we need setfsb to work... unfortunately my vaios pll is not supported, and might actually not support fsb adjustments at all... ill email abo again and will ask him if he can help me out...

[saaya]

i asked abo and franck and they both offered to have a look

i found something interesting in the PLL whitepapers...



one thing that i didnt mention is when we oc the fsb we change the mem clocks as well... keep that in mind... on my vaio the default ratio is 2:3 so at 266fsb i get 400mhz= ddr2 800. if i clock the fsb to 333 im at ddr2 1000 and at 400 im at ddr2 1200 even...

EDIT: in theory, i didnt manage to oc the fsb a single mhz so far

[saaya]

bad news, the bsel modification causing the low multi is probably done on purpose by intel to prevent bsel mods...
at the same time they apparently locked plls on all new laptops to prevent overclocking in a "secure mode" and the pll can only be adjusted within a very short preiod during initialization... so basically only the bios can change the pll clock, if it doesnt, then theres no other way... it feels like we are moving backwards tech wise... i7 is completely locked once it boots up, we cant even change memory timings, and plls and other registers are locked as well... sigh...

ill see if i can find a hw mod to keep either the multi high with a bsel mod or oc via the pll, but i dont have big hopes...
did you try to use crystal cpuid to raise the multi in windows after you did the bsel mod?

[cirthix]

I wonder if the BSEL pins are multipurpose, with state checked at power-up, then monitored after bootup. Maybe instead of removing a pin to set a 1, a weak pullup should be used, or the pin reconnected after it's initial state polled.

I have tried bsel mods on yonahs and was locked to 6x as well. I recall that some people had success with yonahs on non-intel chipsets, but those aren't too common anymore.

[cirthix]

intel hasn't had pin-based multis since the p3 days (if even then)...

[saaya]

so you think we should pull a pin high during bootup and then drop it back to normal once the system booted?
hmmmm a bit trciky, but could be done...

[cirthix]

so you think we should pull a pin high during bootup and then drop it back to normal once the system booted?
hmmmm a bit trciky, but could be done...

It's worth a shot.

I have a mobile 965 laptop on the way and can try it in a week and a half. I don't see it as a risky experiment and I have to swap the motherboard in it anyway.

they apparently locked plls on all new laptops to prevent overclocking in a "secure mode" and the pll can only be adjusted within a very short preiod during initialization... so basically only the bios can change the pll clock, if it doesnt, then theres no other way...

Do you have any additional information regarding this change?

[aicjofs]

Didnt mean to abandon this got really busy. I'll check what PLL is used(I put everything back together, this is why I should take pictures. haha) and mess with a 333 or 400 sometime this week.

Do you have any additional information regarding this change?

Curious too.

intel hasn't had pin-based multis since the p3 days (if even then)...

I think it was the P3 that started the lock.

did you try to use crystal cpuid to raise the multi in windows after you did the bsel mod?

Yes, there was no change. A side note about RMClock, I am unable to disable EIST, I thought I could/did. It says this is the RMClock help file.

"NOTE: some processors have this setting as a read-only - i.e., trying to disable the enabled setting or enabling the disabled won't have any effect. When this case is detected, this option is grayed out. "

Not sure where they got that information but it is indeed greyed out. It might not be that's is read only as much as the CPU is not detected properly but something to note.

[cirthix]

"NOTE: some processors have this setting as a read-only - i.e., trying to disable the enabled setting or enabling the disabled won't have any effect. When this case is detected, this option is grayed out. "

Not sure where they got that information but it is indeed greyed out. It might not be that's is read only as much as the CPU is not detected properly but something to note.

The pinmod works fine on cpus that don't support EIST, like the celeron-m chips.

Disabling EIST in bios on chips that support it also locks them to 6x, in the same way that a detected BSEL mod does.

[ownage]

I can't find any clear mod for my t2310
I would like to boost it from 133fsb to 166fsb. It should be identical to penryn.
Anyone know?

[aicjofs]

I can't find any clear mod for my t2310
I would like to boost it from 133fsb to 166fsb. It should be identical to penryn.
Anyone know?

This would logically be it. Bring BSEL[1] high. Again I haven't heard of not getting the 6x stuck of socket P with Intel chipset.

[saaya]

Do you have any additional information regarding this change?

not really... abo just told me the pll is in secure mode=locked... and in the plls whitepapers it mentions a lot about intel vpro stuff... which i didnt know p45 had integrated already... so apparently intel is using some sort of vanderpool on their laptops? :/

and no, p3 didnt start the lock i think, p4m could still be oced, bsel became popular with northwood and later pentium m cpus... it started not working right with core and core2 iirc...

The pinmod works fine on cpus that don't support EIST, like the celeron-m chips.

Disabling EIST in bios on chips that support it also locks them to 6x, in the same way that a detected BSEL mod does.

so eist has def to do with it... sounds like only bios can fix this since the registers are probably all read only, at least after initialization... damn... why is intel doing this???
to prevent some few hundred geeks like us from ocing their cpus? way to go intel!

so if we are locked to 6x, but can get a bsel working to 400, then thats 6x400=2400mhz... that would still be an oc for most people, and higher fsb means notably better performance...

but what we really need is pll access... how about adjusting the input clocksignal to the pll? that would work, but it would alter ALL clock frequencies... even usb... and ocing the usb ref clock ... hah... thats a nightmare... so in the end the only way is hacking the bios or... modifying the pll in hardware... or even replacing the pll? meh... way to go intel... way to go...

[cirthix]

I just thought of something to test the hypothesis of multipurpose pins VS locked states at bootup:

Attempt to pinmod a cpu to it's stock frequency. IE, take a 166fsb chip and try to pinmod it to 166fsb. Take the low bsels and attach them to nearby ground pins and isolate the normally high bsel pins.

You wouldn't be changing the stock frequency, but you would be effectively breaking the pins for alternative uses. I'm curious to see if this would trigger the same multiplier lock.



another thing to note: modifying VID pins does not trigger the multiplier lock, and these pins aren't directly connected to the VRM on my t60, they are under bios-control (unfortunately not user-selectable, but when disabling EIST to force 6x mode, the cpu gets .95v even if the vid pins have been modified to have the 'stock' voltage be something other than .95v for all multis)

@saaya, a pll mod would depend on the pll used. I used this method to force a t42 (i855 pentium-m, not 915 that supported 133) to 133fsb a long time ago, and it worked great. I tried to do the same sort of mod on my t60 after finding some datasheets of similar clock generators, but the system wouldn't post, I had to revert it back to stock. This may be for other reasons though.

[aicjofs]

Well the mod for 333 and 400 resulted in a no POST condition. Someone feel free to try though maybe I didn't have good connections, the right BIOS, etc.

The PLL clockgenerator is the SiLego slg8sp510 specifically SLG8SP510T this is for the Toshiba L305-S5941 (in case someone is googling for it)

Whitepaper here http://www.silego.com/resources/pdf/...0_07242007.pdf some interesting stuff on the registers too.

Also of note is I changed the SPD of the RAM a while back to 5-5-5 at 400Mhz from 6-6-6, perhaps it is still working at default divider and it's too much for RAM. Maybe the chipset doesn't get enough voltage too, would have to read up on that and make comparisons.

Pictures for informational purposes.

[cirthix]

my t61 board came doa

@aicjofs, try modifiying the clock generator, not the socket.

Also, the lack of post when using the quad-core may be because the bios doesn't support the cpu. with proper bios cede, it might work at reduced speed.

[ziddey]

Sweet, a recent thread on pinmods. This is great, although the news is bad.

So we know yonahs and penryns with intel chipsets don't take kindly to bsel mods. And it's in the chipset itself, so it's probably safe to assume the pentium dual core line (merom and penryn) both won't keep the multiplier free. I've got a t2370 merom and x3100 (960gl?) chipset that I was thinking of playing with. Guess it's pretty pointless now.

But just to confirm, we can use vcc to pull up to logic high? Good to hear.

Also, it's pointless to have a switched bsel, since it's only applied at bootup. Although... if what you're saying is to use a high fsb bsel to bootup, and then switch it back to default fsb bsel, then maybe it'd free the multiplier? I'm guessing it's probably not true, and a lot of work to accomplish. Would more require soldering on the other side of the pcb (assuming it goes through all the layers of pcb?). But if it works, I may just be interested in doing it.

Guess I'll be waiting for used 800fsb merom/penryn parts to get cheaper to upgrade my t2370. Still doing more than adequately for what I need a notebook to do. Extra fsb would be nice though. 133fsb is stoneaged.

[karbonkid]

Someone mentioned they got around this in Linux...need to look into that next.

Maybe cpufreq still works to change multi.. Was this 'someone' a reliable source? You could try a basic linux install or even livecd and play around with specifying different cpufreq drivers (eg. generic ACPI and the various Speedstep specific ones) in the relevant /etc/rc.d startup scripts (which one will vary slightly by distro). You can write performance modes to /sys/devices/cpu/cpu0/cpufreq/scaling-governor, eg. performance (highest possible multi), ondemand etc.

If you need more specific instructions, please ask

I have a dothan in this laptop using 'ondemand' cpufreq governor with the 'centrino' driver and it switches multis on the fly no problem. I think it also worked with the 'acpi' driver. Not sure if any of this is relevant, forgive me if it isn't.

[ziddey]

dothans have no problem with bsel mods. I've read on several occasions of people being able to force multipliers in linux, so it sounds reliable enough.

too bad with meroms. I'm pretty much completely sure that my merom (t2370, 13x133=1733) would be able to run 13x200=2600 no problem.